Structural basis of transcriptional regulation by the HigA

Molecular Microbiology (2019) 111(6), 1449–1462 doi:10.1111/mmi.14229First published online 1 April 2019

Structural basis of transcriptional regulation by the HigA antitoxin

Marc A. Schureck,1†,‡ Jeffrey Meisner,1‡ Eric D. Hoffer,1 Dongxue Wang,1 Nina Onuoha,1 Shein Ei Cho,1 Pete Lollar III2,3 and Christine M. Dunham1*1 Department of Biochemistry, Emory University School of Medicine, Atlanta, GA 30322, USA. 2 Aflac Cancer and Blood Disorders Center, Children’s Healthcare of Atlanta, Atlanta, GA 30322, USA. 3 Department of Pediatrics, Emory University School of Medicine, Atlanta, GA 30322, USA.

Summary

Bacterial toxin–antitoxin systems are important fac-tors implicated in growth inhibition and plasmid maintenance. Type II toxin–antitoxin pairs are regu-lated at the transcriptional level by the antitoxin itself. Here, we examined how the HigA antitoxin reg-ulates the expression of the Proteus vulgaris higBA toxin–antitoxin operon from the Rts1 plasmid. The HigBA complex adopts a unique architecture sug-gesting differences in its regulation as compared to classical type II toxin–antitoxin systems. We find that the C-terminus of the HigA antitoxin is required for dimerization and transcriptional repression. Further, the HigA structure reveals that the C terminus is ordered and does not transition between disorder-to-order states upon toxin binding. HigA residue Arg40 recognizes a TpG dinucleotide in higO2, an evolutionary conserved mode of recognition among prokaryotic and eukaryotic transcription factors. Comparison of the HigBA and HigA-higO2 structures reveals the distance between helix-turn-helix motifs of each HigA monomer increases by ~4 Å in order to bind to higO2. Consistent with these data, HigBA binding to each operator is twofold less tight than HigA alone. Together, these data show the HigB toxin does not act as a co-repressor suggesting potential novel regulation in this toxin–antitoxin system.

Introduction

Nearly all bacteria encode a diverse set of toxin–anti-toxin gene pairs with the capacity to limit cell growth in response to environmental stress (Moyed and Bertrand, 1983; Gerdes et al., 2005; Gonzalez Barrios et al., 2006; Harrison et al., 2009; Wang et al., 2011). These systems consist of a toxin that inhibits cell growth by blocking essential cellular processes during stress and an antitoxin that neutralizes the toxin during normal growth. Under conditions of stress, the labile antitoxins are proteolyti-cally degraded and the more stable toxins are liberated to inhibit growth (Van Melderen et al., 1994; Christensen et al., 2004). In order for toxin–antitoxins to function as effective governors of cell growth in response to stress, several system parameters must be regulated. Prior to the appearance of stress, a sufficient amount of antitoxin needs to be expressed to maintain toxin quiescence, allowing for normal growth. When stress is encountered, toxins need to be freed from their cognate antitoxins to enable cells to halt their growth (through toxin activity). Since toxins are activated by antitoxin proteolysis, too much antitoxin may dampen the responsiveness of the system and delay growth inhibition. Moreover, too much toxin (an amount that exceeds the capacity of antitoxin neutralization) also needs to be avoided, as this would slow or halt growth in the absence of stress. Therefore, the expression of toxin and antitoxins must be finely regu-lated to respond to stress.

To establish and maintain the optimal levels of their two system components, bacterial antitoxins typically serve as transcriptional autorepressors that constitute a negative feedback loop (Page and Peti, 2016). When the appro-priate amount of antitoxin is reached, the antitoxin binds upstream of the toxin–antitoxin operon and represses new transcription. Antitoxins can weakly repress transcription in the absence of their toxin partners ensuring contin-ued transcription until the appropriate amount of toxin is reached (Afif et al., 2001; Overgaard et al., 2008; Garcia-Pino et al., 2010). Often the toxin can serve as a transcrip-tional co- and anti-repressor depending on its abundance relative to the antitoxin (Afif et al., 2001; Overgaard et al., 2008; Garcia-Pino et al., 2010). Once a particular ratio of the toxin and antitoxin is exceeded, the toxin functions as an anti-repressor that severely decreases the anti-toxin affinity for DNA allowing for transcription. For each

© 2019 John Wiley & Sons Ltd

Accepted 16 February, 2019. *For correspondence. E-mail: [email protected] (C.M. Dunham). Tel. 404-712-1756; Fax 404-727-2738. Present address: † The Laboratory of Malaria and Vector Research, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Rockville, MD 20852, USA. ‡These authors contributed equally to this work.

mailto:

mailto:

mailto:[email protected]

mailto:[email protected]

1450 M. A. Schureck et al.

© 2019 John Wiley & Sons Ltd, Molecular Microbiology, 111, 1449–1462

toxin–antitoxin transcript, a disproportionate amount of antitoxin is translated because of the presence of a stron-ger ribosome binding site as compared to the toxin (Li et al., 2014). In some well-studied toxin–antitoxin systems, toxins exert their co-repressor functions by influencing the structure of their antitoxins. Toxin binding can order intrin-sically disordered regions of their antitoxins (Kamada et al., 2003; Garcia-Pino et al., 2010) and can promote and/or disrupt the formation of antitoxin complexes that are proficient for DNA binding (Afif et al., 2001; Kedzierska et al., 2007; Garcia-Pino et al., 2008; De Jonge et al., 2009; Garcia-Pino et al., 2010). Whether these regula-tory mechanisms can be generally applied to understand other toxins-antitoxin systems is unknown.

Antitoxin proteins have two major functions during nor-mal growth. They inhibit toxin function by direct binding and repress transcription to limit expression (Loris and Garcia-Pino, 2014; Page and Peti, 2016). Antitoxins are structurally diverse and contain either a ribbon-helix-helix (RHH) (De Jonge et al., 2009; Boggild et al., 2012), Phd/YefM (Garcia-Pino et al., 2010), SpoVT/AbrB (Dienemann et al., 2011), or a helix-turn-helix (HTH) DNA-binding motif (Brown et al., 2009; Schumacher et al., 2009; Schureck et al., 2014). These distinct antitoxin architec-tures raise the possibility that the regulatory mechanisms that establish and maintain the appropriate amounts of antitoxin proteins may be distinct from those that have been established for more traditional systems including CcdAB, RelBE and bacteriophage P1 Phd-Doc (Afif et al., 2001; Overgaard et al., 2008; Garcia-Pino et al., 2010).

The host inhibition of growth BA (higBA) operon was identified on the Rts1 plasmid associated with Proteus vulgaris in a post-operative urinary tract infection that dis-played resistance to antibiotics (Tian et al., 1996). higBA is classified as a bacterial toxin–antitoxin pair whereby the HigB toxin protein inhibits protein synthesis by cleaving adenosine-rich mRNA transcripts on actively translat-ing ribosomes (Hurley and Woychik, 2009; Schureck et al., 2015; 2016a; 2016b). Homologs of HigBA are also chromosomally distributed although their structure and regulation suggest their mechanism may be distinct from P. vulgaris HigBA (Christensen-Dalsgaard et al., 2010; Kirkpatrick et al., 2016; Hadzi et al., 2017). Structures of the heterotetrameric P. vulgaris HigBA complex (two HigAs and two HigBs) reveal that the HigB toxin is a member of the RelE family of bacterial toxins while the HigA antitoxin contains a HTH DNA-binding motif that recognizes its operator for transcriptional autorepression (Schureck et al., 2014). The structure also indicated a number of differences between the HigBA complex and other RelE toxin–antitoxin family members. For example, other antitoxins that recognize RelE toxin family members typically contain RHH DNA-binding motifs that require dimerization to form a single DNA-binding motif (Kamada

and Hanaoka, 2005; Boggild et al., 2012; Ruangprasert et al., 2014). HigA antitoxin contains a HTH motif that is a complete DNA-binding motif; therefore, the two HigA antitoxins in the HigBA complex contain two DNA-binding motifs in contrast to other RelE family members that contain a single DNA-binding motif. Another distinction of the P. vulgaris HigBA complex is that the HigA anti-toxin does not wrap around the HigB toxin for inactivation but, instead, has a C-terminal extension that engages the adjacent HigA. We previously determined that this C-terminus is critical for HigA dimerization and the HigA dimer is required for binding to its operator (Schureck et al., 2014). In this study, we sought to understand the molecular basis of HigA interaction with its DNA operator.

Results

Dimerization is critical for HigA to repress transcription of hig

The Rts1 plasmid encodes the toxin higB gene followed by the antitoxin higA antitoxin gene, an inverted gene organization as compared to other RelE family members (Fig. 1A) (Gerdes et al., 2005). The hig promoter (Phig) controls the transcription of both genes and is regulated by two operator sites (higO1 and higO2) that overlap with portions of the Phig −35 and −10 promoter sites. The HigA protein binds both operators and represses transcription from Phig. We previously purified a HigA lacking its C-terminal 21 amino acids (∆84-104) that dis-rupts HigA dimerization in the context of the HigBA com-plex (Schureck et al., 2014). Although each monomeric HigA variant contained a complete HTH DNA-binding motif, the HigA monomer was unable to bind to its DNA operator as assessed by electrophoretic mobility shift assay (EMSA). Additionally, disruption of the HigA dimer had no impact on the ability of a monomer of HigB to bind in vitro. We next wanted to test whether the HigA C-terminal truncation specifically disrupted the transcrip-tional repression function of HigA in vivo while preserv-ing its toxin neutralization function. The toxin higB gene, with or without either the full-length or truncated antitoxin higA gene, was placed under the control of the arabi-nose-inducible expression pBAD promoter. The expres-sion plasmids were introduced into E. coli BW25113 and grown to mid-exponential growth phase in rich medium, serially diluted and spotted onto agar plate with or with-out 0.2% arabinose (to induce gene expression). Cell growth was inhibited when expression of the higB toxin gene was induced in the absence of higA and normal cell growth was restored when HigB and full-length HigA were co-expressed, as previously observed (Tian et al., 1996) (Fig. 1B). Importantly, normal cell growth was also observed when HigB and the truncated HigA were

Structural basis of transcriptional regulation 1451


co-expressed, indicating that the C-terminal portion of HigA is dispensable for its toxin neutralization function in vivo (Fig. 1B). Thus, dimerization appears to be feature of HigA that is dedicated to its transcriptional repression function rather than toxin neutralization.

To better understand how HigA controls higBA gene expression in vivo, we constructed a series of transcrip-tional reporters for the hig promoter. First, the hig promoter was cloned upstream of the lacZ gene (encoding the β-ga-lactosidase enzyme) in the transcriptional reporter plasmid pQF50. Next, either the full-length or truncated higA gene (∆84-104) was cloned between the hig promoter and the lacZ reporter gene, creating two different synthetic bicis-tronic operons where both higA and lacZ are expressed under the control of the hig promoter. The transcriptional reporter plasmids were introduced into E. coli BW25113 and grown to mid-exponential phase in rich medium and β-gal activity was measured (Fig. 1C). The strain contain-ing the transcriptional reporter without higA (Phig-lacZ) produced ~600 Miller units (MU) of β-gal activity (Fig. 1C). In the strain containing the transcriptional reporter with the full-length higA (Phig-higA-lacZ), this activity was decreased nearly 60-fold (from ~630 to ~10 MU) through the transcriptional repression function of HigA. The strain bearing the transcriptional reporter with higA(∆84-104) that disrupts HigA dimerization (Phig-higA∆84-104-lacZ), however, had nearly the same amount of β-gal activity as the reporter without higA (~590 MU) (Fig. 1C). One pos-sible interpretation could be that HigA(∆84-104) expres-sion is compromised. However, since coexpression of this variant with HigB leads to normal growth (Fig. 1B), this

strongly suggests that HigA(∆84-104) expression is suf-ficient to suppress HigB toxicity. Collectively, these data demonstrate that dimerization of HigA is necessary for transcriptional repression.

In the pQF50-Phig-higA-lacZ transcriptional reporter plasmid, the level of higA gene expression is a product of the strength of the hig promoter and HigA-mediated transcriptional repression. The configuration of this reporter system is simple and convenient but it lacks the ability to compare the transcriptional repression of HigA variants at defined expression levels. To perform this analysis, either the full-length or truncated higA gene (∆84-104) was cloned into the arabinose-inducible expression plasmid pBAD33. The expression plasmids were then introduced into E. coli BW25113 bearing the Phig-lacZ transcriptional reporter (without higA). The strains were grown to mid-exponential phase in mini-mal medium (M9 maltose) supplemented with 0.2% arabinose and β-gal activity was measured. The strain with the Phig-lacZ transcriptional reporter and pBAD33 (without higA) produced ~4200 MU of β-gal activity (Fig. 1D). The β-gal activity in the strain containing pBAD33 with the full-length higA was decreased more than 20-fold (~150 MU), demonstrating that the amount of HigA produced at this expression level could effectively repress transcription from the hig promoter in trans (Fig. 1D). The strain containing pBAD33 with the truncated HigA(∆84-104) was incapable of high levels of transcrip-tional repression as assessed by its high level of β-gal activity (~3000 MU). These results show that, at defined HigA expression levels, dimerization is necessary for

Fig. 1. HigA dimerization is dispensable for toxin inhibition but necessary for transcriptional repression. A. Organization of the hig operon with the regions HigA recognizes shown in grey shading and the −35 and −10 promoter regions boxed. B. Spot dilution assay of E. coli BW25113 transformed with indicated plasmids, overexpressed and the indicated amounts were plated on LB and LB in the presence of 0.2% arabinose. C. β-gal assays of E. coli BW25113 transformed with indicated plasmids in LB medium. D. β-gal assays of E. coli BW25113 transformed with indicated plasmids in M9 maltose media and 0.2% arabinose.



HigA to repress transcription activity from the hig pro-moter. Additionally, the results show that HigA is capa-ble of repressing transcription from the hig promoter in the absence of HigB, suggesting HigB does not func-tion as a co-repressor in this system. This seems to be in contrast to prevailing models used to describe the transcriptional control of other toxin–antitoxin systems where the toxin is often necessary for effective tran-scriptional repression by the antitoxin (Afif et al., 2001; Overgaard et al., 2008; Garcia-Pino et al., 2010). This is not a trivial distinction between these systems, as con-ditional toxin co- and anti-repressor function is the cen-tral tenant used to describe how toxin–antitoxin systems control the expression of their own genes in order to be primed to respond to stress. These data strongly sug-gest that the molecular mechanisms that underpin the priming of HigBA are distinct from other systems includ-ing RelBE, Phd-Doc, CcdAB, MqsRA, DinJ-YafQ and HicAB (Afif et al., 2001; Overgaard et al., 2008; Garcia-Pino et al., 2010; Brown et al., 2013; Ruangprasert et al., 2014; Turnbull and Gerdes, 2017).

Distance between HigA HTH motifs increases in the absence of HigB

The X-ray crystal structure of the heterotetrameric HigBA complex revealed how the antitoxin HigA inhib-its HigB but whether the toxin HigB influences the over-all conformation of HigA was unclear (Schureck et al.,

2014). This is important because in other toxin–anti-toxin systems, the toxin can profoundly influence the ordering of antitoxin regions that are intrinsically dis-ordered (Kamada and Hanaoka, 2005; Li et al., 2008; De Jonge et al., 2009; Garcia-Pino et al., 2010). This disordered-to-ordered transition has been shown to be important for stabilization of the antitoxin for tran-scriptional repression (Loris and Garcia-Pino, 2014). In the heterotetrameric HigBA complex, the last 11 C-terminal residues of HigA (amino acids 94-104) are disordered in the presence of HigB but this region does not interact with HigB (Schureck et al., 2014). We next solved the X-ray crystal structure of antitoxin HigA in the absence of HigB to 1.9 Å (Fig. 2A; Table S1). HigA crystallizes in the C2 space group with one HigA dimer per asymmetric unit. The final HigA model was built for residues 3–94 and 6–99 (out of 104 total residues) for monomers A and B of the dimer, respectively. HigA is a single domain protein and adopts a compact, five α-helix bundle with α2 and α3 comprising the HTH motif that is exposed at its N-terminus (Fig. 2A). HigA dimerization is mediated via interactions between α5 and its C-termini that extend to interact with each HigA monomer. Recognition of the hig operator is likely mediated by α3 while α2 helps to position α3 in the major groove consistent with other classical bacterio-phage transcriptional repressors that HigA structurally resembles (Shimon and Harrison, 1993; Watkins et al., 2008). These bacteriophage repressors form dimers

Fig. 2. HigA is an obligate dimer. A. 1.9 Å X-ray crystal structure of HigA reveals the maintenance of the dimer interface in the absence of the HigB toxin. The DNA binding helix-turn-helix (HTH) motif and the dimer interface are indicated. B. Analytical ultracentrifugation of HigA produced a signal-average s20,w peak at 2.15 S corresponding to an estimated molecular weight of 25.9 kDa. C. Comparison of the HigA dimer (green) and HigA in the context of the HigBA complex (blue; PDB code 4MCT) reveals a 5° move away from the DNA binding surface involving α1, α2 and α3.



that separate the recognition helices, one from each monomer, by the same distance as successive major grooves of the operator DNA. Sedimentation velocity analytical ultracentrifugation (AUC) studies show that HigA is a 2.15 S particle with an estimated molecular mass of 25.9 kDa (the molecular mass of each HigA polypeptide is 11.5 kDa) providing additional support that HigA is a dimer (Fig. 2B).

The overall structural architecture of the HigA dimer remains largely unchanged in the presence or absence of the toxin HigB (root mean square deviation (rmsd) of ~1.1 Å for 185–186 α carbon backbone atoms aligned). This comparison provides support that the toxin HigB does not induce large conformational rearrangements of HigA upon binding in contrast to other systems as men-tioned above. However, one noticeable change is seen at the HigA dimer interface (Fig. 2C). Aligning a single protomer of the HigA dimer reveals a movement of HTH α-helices 1, 2 and 3 backbone residues of ~4.0 Å and a shift outward of 5.1° in the absence of toxin HigB (Fig. 2C). This movement increases the distance between each HTH motif by ~1.5 Å (to 29.4 Å from 27.9 Å as measured from Arg40 in each protomer) as compared to the HigA dimer in the HigBA complex (Schureck et al., 2014). If the spacing between α3 of the two HigA protomers is import-ant for recognition of the hig operator, the remodeling of the HigA dimeric interface induced by HigB binding could modulate the strength of HigB binding to hig.

Structure of HigA bound to higO2

To understand how HigA recognizes its operator at the molecular level, we solved the X-ray crystal structure of HigA bound to the higO2 DNA operator to 2.9 Å (Fig. 3; Table S1). The HigA-DNA complex crystallizes in the space group P65 with three HigA dimers and three 21 nucleotide, double-stranded DNA in the asymmetric unit. The final HigA model was built for residues 4–92 and 4–91 (out of 104 total residues) in dimer A, 5–92 and 5–95 for dimer B and 6–96 and 7–92 for dimer C. The three HigA dimers in the asymmetric unit are similar with rmsd values ranging 0.424–0.537 Å. A DALI search reveals that HigA-DNA is similar to the Shewanella onei-densis HipA-HipB-DNA toxin–antitoxin complex (PDB code 4PU4), Pseudomonas aeruginosa transcriptional repressor RsaL-DNA complex (PDB code 5J2Y) and Escherichia coli HipB antitoxin-DNA complex (PDB code 4YG1) with Z-scores of 6.2, 5.8 and 5.3, and r.m.s.d. values of 5.5, 2.1 and 2.9 Å, respectively (using 66, 55 and 59 aligned Cα atoms, respectively) (Holm and Rosenstrom, 2010).

Each HigA dimer recognizes a single inverted repeat of the operator with no apparent crosstalk between HigA

dimers (Fig. 3A). Sequence-specific interactions are medi-ated by α3 recognition of the nucleobases in the major groove while sequence-independent interactions with the phosphate backbone on the opposite strand are mediated by the α1–α2, α2–α3 and α3–α4 loops. As is often the case with HTH motifs (Shimon and Harrison, 1993), three residues (Thr34, Thr36 and Arg40) are displayed on the surface of the α3 recognition helix facing the major groove of the inverted operator repeats (Fig. 3B). The hydroxyl groups of Thr34 and Thr37 and the backbone amine of Thr34 interact with the G+7 phosphate while the side-chain methyl groups of Ala36 and Thr34 form van der Waals interactions with the nucleobase C5 methyl of T+6 (Fig. 3C). HigA also makes hydrophobic interactions in recognition of the operator via the side-chain methyl group of Thr46 with the C5 methyl of T+8. Sequence-specific interactions include hydrogen bonds between the hydroxyl groups of Ser23 and Ser39 with the G+12 and T+10 phosphate oxy-gens, respectively, and an ionic interaction between the side chain amino group of Lys45 and the T+9 phosphate oxygen. Non-sequence-specific interactions between HigA and its operator are between the backbone amines of both Gly24 and Arg25 and the phosphate oxygens of A+11.

HigA α3 residue Arg40 specifically recognizes the DNA major groove. The Arg40 side-chain guanidino moiety makes bifurcated hydrogen bonding interactions with the Hoogsteen face of G + 7 (via the O6 and N7 atoms) (Fig. 3B). Additionally, the guanidino moiety stabilizes interac-tions with the C5 methyl group of T+8 via cation-pi inter-actions. The interaction between Arg40 and the T+8G+7 dinucleotide is similar to an evolutionary conserved mode of recognition of TG and methylated CG dinucle-otides by prokaryotic and eukaryotic transcription factors termed YpG interactions (where Y indicates a pyrimidine) (Lamoureux et al., 2004; Liu et al., 2013). Together, these interactions likely facilitate HigA specific recognition of higO2.

To identify how HigA binding might alter higO2, we analyzed the backbone and nucleobase geometry of higO2 DNA using the program Curves+ (Blanchet et al., 2011). The most prominent finding of this analysis was that there are local deformations of the major and minor grooves as compared to B-form DNA. In the structure of HigA bound to higO2, the major groove was narrowed at the site where the HTH inserts (~9.9–10 Å) and widens (~11.8 Å) between both HTH motifs on the opposite side of the DNA (Fig. S1) (the major groove of B-form DNA is 11.4 Å). Likewise, the minor groove expands (~7.4–7.7 Å) on the opposite side of where the HTH motifs contact DNA and contracts in the center of higO2 between the HigA monomers (~6.9 Å). These backbone deformations from B-form DNA may be required to facilitate productive HigA–DNA interactions (Fig. S1).



HigA arginine 40 governs hig recognition and transcriptional repression

To investigate the importance of Arg40 in operator rec-ognition, we constructed a HigA variant with an Arg40 to Ala substitution and analyzed its function both in vivo

and in vitro. To confirm that Arg40 is important for the transcriptional repression function of HigA and not its toxin neutralization function, we performed the spot-di-lution assay described in Fig. 1C to evaluate the ability of the point mutant to inhibit HigB toxicity. Normal cell

Fig. 3. Structural basis of HigA-DNA operator recognition. A. 2.9 Å X-ray crystal structure of HigA bound to higO2. The higO2 is shown with the blue arrows indicating the inverted repeats HigA recognizes and the specific nucleotides contacted are shown in bold. B. Zoomed in view of HigA α2 and α3 of the HTH motif that interact directly with nucleotides T+6, G+7 and T+8. HigA residue Arg40 hydrogen bonds to G+7. The phosphate of G+7 is contacted by Thr34 and Thr37 that may serve to stabilize the interaction between Arg40 and the nucleobase of G+7 (right panel). C. Schematic representation of interactions between HigA residues with higO2. Van der Waals interactions between Ala36 in both monomers are shaded grey.



growth was observed when HigB and the HigA R40A variants were co-expressed, indicating that HigA Arg40 was dispensable for its toxin neutralization function in vivo (Fig. 4A). To examine the importance of Arg40 in the ability of HigA to control higB-higA transcription, a pBAD33 plasmid with the HigA R40A variant was intro-duced into E. coli BW25113 bearing the Phig-lacZ tran-scriptional reporter plasmid. When the strain was grown to mid-exponential phase in minimal medium with 0.2% arabinose, it produced nearly the same amount of β-gal activity as the strain with pBAD33 without higA (Fig. 4B). This result demonstrates that Arg40 is important for HigA to repress transcription from the hig promoter. We next determined that HigA R40A is unable to bind to its DNA operator in vitro, even up to a concentration of 1 μM HigA R40A protein (Fig. 4C). Further, we confirmed that

HigA R40A retained its ability to form a dimer (molecu-lar weight of ~30 kDa), an oligomeric state known to be required for DNA binding in vitro (Schureck et al., 2014) (Fig. 4D). Lastly, we mutated both guanosines in each inverted repeat to adenosines (G+7 as shown in Fig. 4B) and observed no wild-type HigA binding (again to a con-centration of 1 mM) (Fig. 4E). These data demonstrate that HigA Arg40 is important for operator recognition and interacts specifically with nucleotide G+7 but is not critical for HigA dimerization or HigB neutralization.

Comparison of HigBA, HigA and HigA-higO2 structures

Although the HigB toxin is not required for the forma-tion of the HigA dimer (Schureck et al., 2014), binding of HigB does influence the overall architecture of HigA as evidenced by superimposition (Fig. 5A). The dis-tance between the HigA HTH motifs located in each protomer when bound to HigB is 27.2 Å (as measured by the distance between Arg40 located in the middle of each operator recognition helix α3 of each promoter). HigA binding to higO2 increases the distance between the HTH motifs to by ~4 Å to 31.0 Å and moves a HigA promoter 12° relative to the other promoter. This move-ment is necessary to expand the distance between the two recognition helices of the HigA dimer to bind to the two inverted repeat sequences and two major grooves of the operator. In the presence of HigB, the distance between the HTH motifs decreases to a distance that would cause a steric clash between HigA helix α3 and higO2 (Fig. 5A).

These structural comparisons also suggest other changes to HigA that may occur upon HigB binding (Fig. 5B). For example, in the presence of the toxin HigB but the absence of higO2, the N-terminal loop of HigA packs against HigB and forms a hydrogen bond and salt bridge network between HigA residues Arg2 and Gln3 with Glu80 and Arg77 of HigA α5, respectively (Fig. 5B, top). In the HigA dimer structure solved in this study, these interactions between HigA loop 1 and α5 are weakened by rotation of α5 residues that ablate and remodel these interactions (Fig. 5B, bottom). This disruption between Arg2 and Glu80 is also seen in the HigA-higO2 oper-ator structure. Considering that the HigA dimerization occurs at the α5–α5 interface and remodeling of the HigA protomers relative to the other in the absence of HigB appears to be mediated by the α5–α5 interface (Fig. 2A and C), the consequence of disrupting interac-tions between HigA loop 1 and α5 may contribute to the increase in distance between HigA HTH recognition heli-ces (Fig. 5A). These structural comparisons suggest that HigB may help to rigidify HigA’s conformational freedom of rotation.

Fig. 4. HigA Arg40 and G+7 are necessary for HigA recognition of higO2. A. Spot dilution assay of E. coli BW25113 transformed with indicated plasmids, overexpressed with the indicated amounts were plated on LB and LB in the presence of 0.2% arabinose. B. β-gal assays of E. coli BW25113 transformed with indicated plasmids. C. Electrophoretic mobility shift assay of HigA R40A and higO2. D. Size exclusion chromatography analysis of HigA wild-type and the R40A variant. E. Electrophoretic mobility shift assay of the wild-type HigA and higO2 containing a G+7–A+7 mutation (along with a C+7 to U+7 to maintain Watson-Crick base-pairing). [Colour figure can be viewed at wileyonlinelibrary.com]

www.wileyonlinelibrary.com



The HigBA complex binds twofold less tight than HigA to higO1 or higO2

Previously, we and others have found that the HigBA complex binds DNA (Tian et al., 2001; Schureck et al., 2014). However, it is unclear how the conformational changes observed in HigA alone versus in complex with HigB will affect operator binding. Therefore, we next per-formed EMSAs to assess the ability of purified HigA or HigBA to bind to a 61 base-pair DNA fragments that con-tained higO1 with higO2 sequence scrambled or higO2 with higO1 sequence scrambled (Figs. 6 and S2). HigA binds to each of operator with approximately the same affinity (Kd values of 140 ± 0.03 and 125 ± 0.03 nM for higO1 with higO2, respectively). Interestingly, HigBA

binds to each operator with a ~twofold lower affinity (Kd values of 364 ± 0.10 and 243 ± 0.04 nM for higO1 with higO2 respectively) (Figs. 6B and S2).

Discussion

Much of what is known about the transcriptional regula-tory mechanisms that establish and maintain appropriate levels of toxin–antitoxin complexes has been derived from the careful examination of a small set of model systems (reviewed in (Loris and Garcia-Pino, 2014; Page and Peti, 2016)). Studies of these model systems have led to signifi-cant advances in our understanding of how toxin–antitoxins are regulated but it is unclear whether this understanding

Fig. 6. HigA and HigBA recognition of higO1 and higO2. A. The band intensities from EMSAs as plotted as percent HigA bound versus HigA concentration for binding to Phig with scrambled higO1 and of HigA and Phig with scrambled higO2. In each assay, the HigA concentration increases from 0 to 1000 nM (gel shown in Fig. S2). B. The band intensities from EMSAs as plotted as percent HigBA bound versus HigBA concentration for binding to Phig with scrambled higO1 and of HigBA and Phig with scrambled higO2. In each assay, the HigBA concentration increases from 0 to 600 nM (gel shown in Fig. S2). Curves represent the data from which binding affinities given in the main text were derived.

Fig. 5. Structural rearrangements of HigA during operator recognition. A. Superposition of free HigA (this study), the HigB–HigA complex (PDB code 4MCT) and DNA operator-bound HigA (this study) using the second HigA monomer (grey) as an anchor point. The HigA dimer hinges ~12° away from the DNA surface upon DNA recognition as compared to HigA bound to HigB. In the context of both apo HigA and in the HigB-HigA complex, clashes between α2 and α3 with DNA would occur. Rearrangement of the one monomer of HigA is required to allow the HTH motif to fully engage DNA. B. The HigA N terminus packs against the toxin HigB and forms interactions with α5 that may restrict its conformation (top panel). In the absence of HigB, the interactions between loop 1 and α5 are disrupted (bottom panel).



can be broadly applied across diverse classes of toxin–antitoxins. One potential reason for the possibility of different repressor mechanisms is the rich diversity in the structure of antitoxins. Four predominant motifs are exploited by antitoxins for DNA binding: the RHH, Phd/YefM, SpoVT/AbrB or HTH motifs. There are only three known HTH-containing antitoxins and even among this group, there are clear differences in how they regulate tran-scriptional repression. For example, both antitoxins HipB and MqsA cause substantial DNA bending while we find in this study that HigA does not (Schumacher et al., 2009; Brown et al., 2013). Also, the direct and indirect readout of the operator sequences occurs by distinct mechanisms. We demonstrate that HigA recognizes its operator via a mechanism employed by both prokaryotic and eukaryotic transcription factors whereby the side chain of a conserved arginine residue makes both nucleobase-specific interac-tions with a guanosine and pi-cation stacking interactions with an adjacent thymine in recognition of a YpG dinucle-otide (Lamoureux et al., 2004; Liu et al., 2013). In contrast, antitoxins HipB and MqsA do not recognize YpG dinu-cleotides. These data indicate that antitoxins containing HTH DNA-binding motifs do not necessarily use the same mechanism of transcriptional repression.

Previously it has been shown that the transcription of toxin–antitoxin operons can be mediated by higher order toxin–antitoxin complexes, which are regulated by excess toxin functioning as a co-repressor (Afif et al., 2001; Overgaard et al., 2008; Garcia-Pino et al., 2010). For example, when CcdB, RelE or Doc toxins are present at levels below those of their cognate antitoxins, toxin bind-ing to antitoxin-DNA operator sites increases the overall affinity and the level of repression at the toxin–antitoxin operon (Magnuson and Yarmolinsky, 1998; Afif et al., 2001; Garcia-Pino et al., 2010). On a molecular level for the RelBE system which is the most similar to the HigBA system, this high affinity complex has been proposed to be a trimeric toxin–antitoxin complex (two antitoxins and one toxin molecule) when bound to its DNA operator (Overgaard et al., 2008; Boggild et al., 2012). At present, only a structure of the tetrameric RelBE complex has been observed in the absence of DNA (Boggild et al., 2012). As toxin levels increase to those similar to the antitoxin, the additional toxin disrupts the antitoxin-operator complexes and transcriptional repression is relieved (Overgaard et al., 2008). Additionally, in these systems, the antitoxin C terminus is disordered in the absence of its cognate toxin and gains structure upon toxin binding which also leads to an increased binding affinity for its DNA operator. These conditional co- and anti-repressor functions of toxins and their influence on antitoxin structure are the hallmarks of the current models describing toxin–antitoxin regulation.

Structural comparisons of apo HigA, HigA-higO2 com-plex and HigBA provide insights into how transcriptional

repression could be mediated in this system. First, we find that the antitoxin HigA does not undergo a disorder-to-or-der transition at its C terminus in contrast to well-stud-ied toxin–antitoxin systems (Magnuson and Yarmolinsky, 1998; Afif et al., 2001; Overgaard et al., 2008; Garcia-Pino et al., 2010). Instead, we determine that the C terminus of HigA is required for its dimerization and this oligomeric state is, in turn, necessary for DNA operator binding, despite each HigA monomer containing a single DNA operator. Toxin HigB binding does not order HigA provid-ing further support that this system is not regulated by a disorder-to-order mechanism. Lastly, the HigBA complex, that we presume from our previous two crystal structures is a tetrameric complex (two HigBs and two HigAs) in the absence of DNA (Schureck et al., 2014), binds with a ~twofold lower affinity than antitoxin HigA alone. These data demonstrate that toxin HigB does not function as a co-repressor in contrast to a number of other systems. Interestingly, in the case of the antitoxin MqsA, its cog-nate toxin MqsR also does not function as a co-repressor and shares a binding surface on MqsA with its operator DNA, effectively functioning as a competitor (Brown et al., 2013). One formal possibility is that the oligomeric state of HigBA changes in the context of binding the hig promoter, however, further detailed biophysical or structural insights are needed to unravel these molecular details.

These data together provide a possible explanation for how toxin HigB may modulate the ability of the HigA antitoxin to repress transcription of the higBA gene. When HigB is present at levels below those of HigA, HigA binds tightly to its operator sites within the hig promoter and represses expression of higBA genes. When HigB reaches levels similar to those of HigA, the binding of two HigB monomers to the HigA dimer may restrict the complex to a conformation that is less favorable for operator binding than HigA alone. Indeed, structural superimpositions demonstrate that there are clashes of HigA’s DNA-binding motifs in the HigBA com-plex with higO2. Perhaps a single toxin HigB could bind to the HigA dimer which may not result in clashes with higO2, however, at this time, there is no direct evidence that a trimeric HigBA exists. This possible model for how HigB could modulate the ability of HigA to repress gene expression is attractive in that it serves the same operating principles as those invoked to understand the well-studied toxin–antitoxins but does so through a dis-tinct set of molecular mechanisms.

Experimental procedures

General methods

The strains, plasmids and oligonucleotide primers used in this study are listed in Tables S2, S3 and S4 respectively.



Strains were grown in lysogeny broth (LB; 1% tryptone, 0.5% yeast extract, 1% sodium chloride) and M9 minimal medium (48 mM sodium phosphate dibasic, 22 mM potassium phos-phate monobasic, 8.6 mM sodium chloride, 19 mM ammo-nium chloride, 2.0 mM magnesium sulfate, 0.1 mM calcium chloride) with 0.4% maltose. When necessary, strains were grown in media supplemented with 100 μg ml−1 ampicillin, 20 μg ml−1 chloramphenicol or 30 μg ml−1 kanamycin.

Plasmid construction

To build a replicating plasmid with a lacZ transcriptional reporter for the Rts1 hig promoter, oligonucleotides oJM835 and oJM836 were annealed and then ligated to pQF50 that had been cut with PstI and HindIII to make pQF50 Phig-lacZ (pJM359) (Table S3). The pQF50 plas-mid contains a multiple cloning site immediately upstream of a lacZ reporter gene with the strong ribosome-binding site from the E. coli lpp gene. To build a similar replicat-ing plasmid with the higBA and lacZ genes under the con-trol of the hig promoter, a DNA fragment containing the hig promoter, higB and higA was chemically synthesized (IDT). The fragment was cut with PstI and HindIII and ligated to pQF50 cut with the same restriction enzymes to make pQF50-Phig-higBA-lacZ.

To construct a replicating plasmid with higB under the control of an inducible promoter, the higB sequence was amplified by PCR from pQF50-Rts1 Phig-higBA-lacZ with oligonucleotide primers oJM821 and oRJD2. The higB PCR product, as well as the pBAD33 plasmid (Guzman et al., 1995), were cut with PstI and HindIII and then ligated to make pBAD33-higB (pJM346). A series of plasmids were also built containing both higB and higA under the control of the arabinose-inducible promoter in pBAD33. The higB–higA sequence was amplified by PCR from pQF50-Rts1 Phig-higBA-lacZ with oligonucleotide primers oJM821 and oJM837, cut with PstI and HindIII and then ligated to pBAD33 cut with the same restriction enzymes. To build a derivative of this plasmid with the C-terminally truncated HigA, higA∆84-104 was amplified from pQF50-Rts1 Phig-higBA-lacZ with oJM821 and oJM862, cut with PstI and HindIII, and then ligated to pBAD33. To make a derivative with the HigA R40A mutant, pBAD33-higB-higA was subjected to site-directed mutagenesis (according to the Strategene QuikChange pro-tocol) with oJM870 and oJM871. All pBAD33 plasmids were confirmed by sequencing (Genewiz).

To build a transcriptional reporter plasmid with the higA and lacZ genes under the control of the hig promoter, oligonucle-otides oJM863 and oJM864 were annealed and then ligated to pQF50 that had been cut with SphI and BamHI to make pQF50 Phig-lacZ (pJM379). The higA sequence was ampli-fied by PCR from pQF50-Rts1 Phig-higBA-lacZ with oligonu-cleotide primers oJM865 and oJM837. The PCR product was cut with BamHI and HindIII, and then ligated to pJM379 cut with the same restriction enzymes to make pQF50 Phig-higA-lacZ (pJM384). The C-terminally truncated higA, higA∆84-104, was also amplified with oligonucleotide primers oJM865 and oJM837, cut with BamHI and HindIII and then ligated to pJM379 to make pQF50 Phig-higA∆84-104-lacZ (pJM385).

To construct a plasmid with higA under the control of the arabinose inducible promoter, the higA sequence was

amplified by PCR from pQF50-Rts1 Phig-higBA-lacZ with oligonucleotide primers oJM823 and oJM837. The PCR product was cut with PstI and HindIII, and then ligated with pBAD33 cut with the same restriction enzymes. To build a version of this plasmid containing the C-terminally truncated higA, higA∆84-104 was amplified from pQF50-Rts1 Phig-higBA-lacZ with oJM823 and oJM862, cut with PstI and HindIII and then ligated with pBAD33. To make a derivative with the higAR40A mutant, pBAD33-higA was subjected to site-directed mutagenesis with oJM870 and oJM871.

The expression plasmid used in this study, pET28a-his6-Rts1 higA, was a generous gift from the Woychik laboratory (Hurley and Woychik, 2009). To make a derivative of this expression, plasmid with the HigA R40A variant, pET28a-his6-Rts1 higA was subjected to site-directed mutagenesis with oJM870 and oJM871. The plasmids were confirmed by sequencing (Genewiz) using oJM437 and oJM438.

Spot dilution assays

Strains were grown in 3 ml LB supplemented with chloram-phenicol at 37°C with rolling to an OD600 of 0.5–1.0. The cultures were then serially diluted 10-fold in LB, 5 μl of each dilution (10−1–10−6 dilutions) was spotted on LB agar with chloramphenicol or LB agar with chloramphenicol and 0.2% arabinose and cells were grown at 37°C overnight.

β -galactosidase assays

Strains were grown in 3 ml medium (LB or M9 maltose) at 37°C with rolling until cell density reached until an OD600 of 0.5. Cultures were diluted 100-fold into 20 ml medium with or without 0.2% arabinose and grown at 37°C with rolling to an OD600 of ~0.5 (mid-exponential growth phase). About 1 ml of each culture was centrifuged at 10,000 × g in a microcen-trifuge tube for 1 min and cell pellets were stored at −20°C. Cell pellets were thawed on ice and resuspended in 500 μl cold buffer (60 mM sodium phosphate dibasic, 40 mM sodium phosphate monobasic, 10 mM potassium chloride, 1 mM magnesium sulfate, pH 7.0 with 50 mM β-mercaptoethanol). About 10–200 μl of each cell suspension was added to micro-centrifuge tubes containing 800–990 μl buffer (to a final com-bined volume of 1 ml, 100 μl chloroform, and 50 μL 0.1% SDS. Reactions were vortexed and incubated at 30°C for 10 min. 200 μl 4 mg ml−1ortho-nitrophenyl-β-galactosidase (in 0.1 M phosphate buffer, 60 mM sodium phosphate dibasic, 40 mM sodium phosphate monobasic, pH 7.0) was added to each sample. Reactions were vortexed briefly and incubated at 30°C for 10–30 min. Each reaction was terminated by the addition of 400 μL of 1 M sodium carbonate, vortexed and centrifuged to remove cell debris. About 1 ml of each reaction supernatant was transferred to disposable cuvettes. Absorbance of each reaction was measured at 420 nm. β-galactosidase activity (in Miller units) was calculated as follows: (1000 x A420)/(reaction time in minutes x cell suspension volume in ml × OD600).

HigA and HigA variant expression and purification

To overexpress wild-type Rts1 HigA or the HigA R40A variant, the expression strain E. coli BL21(DE3) was



transformed with pET28a-his6-higA or pET28a-his6-hi-gAR40A (pJM382) to make strains BL21(DE3) pET28a-his6-higA (ECJM901) and BL21(DE3) pET28a-his6-higA R40A (ECJM902) respectively. These expression strains were grown in 20 ml LB supplemented with kanamycin at 37°C with shaking. When the cultures reached mid-exponential phase (optical density (OD) at 600 nm of approximately 0.5), they were transferred to a 2.8 L-baffled Fernbach flasks con-taining 1 L LB with kanamycin and incubated at 37°C with shaking. Once the cultures reached an OD600 of 0.6–0.7, they were cooled at 4°C for 15 min. Then 50 μl 1 M isopro-pyl-1-thio-ß-D-galactopyranoside (IPTG) was added to the 1 L cultures to give a final concentration of 50 μM, and the cul-tures were incubated at 18°C with shaking overnight (~18 h). The cultures were then centrifuged, the supernatants were removed and the cell pellets were stored at −80°C.

To purify the wild-type and protein variants, the frozen cell pellets were thawed and resuspended in 25 ml HisTrap buffer A (50 mM Tris, 250 mM KCl, 5 mM MgCl2, 5 mM imid-azole, 5 mM β-mercaptoethanol, 10% glycerol, pH 7.4) with ProBlock Gold protease inhibitor (GoldBio) and 50 μl 10 mg/ml DNase I. The cells were lysed with French press at 15,000 psi. The resulting cell lysates were centrifuged at 10,000 × g at 4°C for 20 min. Cleared lysates were filtered and loaded onto 5 ml HisTrap HP column in an ÅKTApurifier chromatog-raphy system (GE Healthcare). His6-tagged proteins were eluted from the column with an increasing linear gradient of HisTrap buffer B (50 mM Tris, 250 mM KCl, 5 mM MgCl2, 500 mM imidazole, 5 mM β-mercaptoethanol, 10% glycerol, pH 7.4). To remove the hexahistidine tag, HisTrap elution fractions were pooled and incubated with thrombin (1 unit per mg of protein) at room temperature for 2 h and then at 4°C overnight. The thrombin proteolysis reactions were sub-jected to size exclusion chromatography on a HiLoad 16/60 Superdex 200 column and buffer exchanged into storage buffer (40 mM Tris-HCl, pH 7.5, 250 mM KCl, 5 mM MgCl2 and 5 mM β-mercaptoethanol). To remove any remaining thrombin or His-tagged HigA protein, the gel filtration frac-tions were pooled and loaded onto 1 ml HiTrap Benzamidine FF and 1 ml HisTrap FF columns arranged in tandem. Finally, HigA was dialyzed back into storage, concentrated and stored in aliquots at −80°C. The HigA protein lacking any tags is 11.5 kDa.

HigA crystallization, data collection and structure determination

HigA crystals (10 mg ml−1) were grown by sitting drop vapor diffusion in 0.2 M NaSCN and 20% w/v polyethylene glycol (PEG) 3350 at 20°C. Crystals were cryoprotected by grad-ually increasing the ethylene glycol concentration to 30% w/v while decreasing the PEG 3,350 concentration to 10% w/v followed by freezing in liquid nitrogen. Three hundred and sixty degrees of diffraction data were collected at the Northeastern Collaborative Access Team (NE-CAT) beam-line 24-IDC at the Advanced Photon Source (Chicago, IL, USA). The X-ray diffraction data were indexed, integrated and scaled in XDS (Kabsch, 2010). The structure was solved by molecular replacement using a HigA monomer from the HigB-HigA complex (PDB code 4MCT) as a search model in PHENIX AutoMR to 1.9 Å (Adams et al., 2010). Two

HigA proteins per asymmetric unit were identified and form a dimer within the asymmetric unit. The model was built in Coot (Emsley et al., 2010) for residues 3–94 and 6–99 fol-lowed by refinement of xyz coordinates, occupancies and B-factors in PHENIX (Adams et al., 2010) to a final Rwork/Rfree of 17.5/22.2%.

AUC studies of HigA

Purified HigA was dialyzed against 20 mM Tris, 100 mM NaCl, 10 mM MgCl2, pH 7.5. Sedimentation velocity exper-iments were performed on a 0.11 mg/ml sample (0.4 ml) at 182,000 × g (50,000 rpm) at 20°C in a Beckman Coulter ProteomeLab XLI analytical ultracentrifuge using standard procedures (Zhao et al., 2013). Absorbance scans were taken at 280 nm in continuous mode using a radial spacing of 0.003 cm at approximately 3.6 min intervals in an An-60 rotor equipped with 12 mm path length double sector cells and sapphire windows. Absorbance values were fit to the Lamm equation along with the meniscus position, baseline and time-invariant noise using the continuous c(s) distribu-tion model of SEDFIT, version 15.01c (http://analyticalultra-centrifugation.com) integrating between 0 and 10 S at 0.1 S increments (Schuck, 2000; Dam and Schuck, 2004). Fitting was done using maximum entropy regularization with a con-fidence interval of 0.68. Both the simplex and Marquardt-Levenberg nonlinear least squares fitting algorithms were tested and produced equivalent results. Buffer density, vis-cosity and the partial specific volume of HigA (0.730 ml g−1) were estimated using SEDNTERP (http://sednterp.unh.edu) (Cole et al., 2008). The sedimentation coefficient was corrected to s20,w using the buffer density and viscosity. The molecular weight of HigA was obtained from the c(s) peak by deconvolution of the contribution of diffusion to the observed signal as described (Schuck, 2000; Dam and Schuck, 2004). AUC results were plotted using GUSSI ver-sion 1.2.1 (Schuck et al., 2016).

HigA-DNA operator complex crystallization, data collection and structure determination

The hig operator 2 DNA was formed by incubating two 21 nt (pHigCryst3 and pHigCryst4), single-stranded DNA at 95°C for 2 min and then cooled to room temperature for 2 h in DNA buffer (100 mM NaCl, 1 mM EDTA and 10 mM Tris, pH 8.0). HigA (3.2 mg ml−1) was complexed with 1.6 mg ml−1 of the higO2 DNA in 1X binding buffer (20 mM Tris, pH 8, 100 mM NaCl, 10 mM MgCl2) to form a com-plex consisting of 1 HigA dimer per 1 duplex DNA molar ratio. Crystals were grown by sitting drop vapor diffusion in 0.2 M CaCl2 and 10–25% w/v PEG 3,350 at 20°C pro-ducing rod-shaped crystals after two days. Crystals were cryoprotected by three serial increases of ethylene gly-col to final concentration of 30% w/v followed by freezing in liquid nitrogen. The HigA-DNA crystals are of the P65 space group and diffracted anisotropically. A total of 120° of diffraction data was collected on the SER-CAT 22ID beamline. The data were indexed, integrated and scaled to 2.9 Å in XDS (Kabsch, 2010). Structure determination was performed by molecular replacement using the structure

http://analyticalultracentrifugation.com

http://analyticalultracentrifugation.com

http://sednterp.unh.edu

http://sednterp.unh.edu



of HigA in isolation as a search model in PHENIX AutoMR (Adams et al., 2010). Three copies of a HigA dimer bound to double-stranded DNA are found per asymmetric unit. Manual modification of the model was performed in Coot (Emsley et al., 2010) followed by refinement of XYZ coordi-nates with secondary structure restraints and B-factors in PHENIX. The final model has an Rwork/Rfree of 22.5/26.0%. All figures were created in PyMOL (Schrodinger, 2010). Alignments between HigBA, HigA and HigA-DNA were constructed using either the secondary structure matching or least-squares fit functions in Coot (Emsley et al., 2010).

Electromobility shift assay

To construct the double-stranded DNA (dsDNA) for the EMSA, pairs of complementary single-stranded oligonu-cleotides were diluted to 2 μM each in 10 mM Tris, 1 mM EDTA, 100 mM NaCl, pH 8.0. The 61-nt oligonucleotide mix-tures of the hig promoter fragment were incubated at 95°C for 2 min and then cooled at room temperature for 2 h. The wild-type operator was generated from pHigA_F and pHi-gA_R, scrambled operator 1 with wild-type operator 2 from pHigA_F_Scra1 and pHigA_R_Scra1 and wild-type opera-tor 1 and scrambled operator 2 from pHigA_F_Scra2 and pHigA_R_Scra2. For the EMSA, the dsDNA oligos were diluted to 150 nM in EMSA binding buffer (100 mM NaCl, 10 mM MgCl2, 5% glycerol, 0.01 mg/ml BSA). Purified wild-type HigA and HigA R40A proteins were diluted to 10 μM in EMSA binding buffer and serially diluted to give a series of final protein concentrations ranging from 62.5 nM to 1.0 μM in the EMSAs. HigBA was purified as previously described (Schureck et al., 2014) and diluted to 8 μM in EMSA bind-ing buffer and serially diluted to give a series of final pro-tein concentrations ranging from 25 nM to 600 nM in the EMSAs. The binding reactions were incubated on ice for 20 min and 5 μL of each reaction was loaded onto an 8% native, polyacrylamide-0.5X TBE-glycerol gels (50 mM Tris, 50 mM boric acid, 5 mM EDTA, 10% glycerol) and subjected to electrophoresis at 100 volts at 4°C. To visualize the DNA and DNA-protein complexes, the gels were stained with SYBR green nucleic acid gel stain (ThermoFisher Scientific) in 0.5X TBE glycerol for 30 mins with gentle agitation and then the fluorescence was imaged with a Typhoon Trio phosphoimager (GE Healthcare; 488 nm excitation and 526 nm emission). Assays were performed in duplicate with representative gels shown. Band intensities for both free and bound hig DNA were quantified with ImageQuant 1D gel analysis using the rolling ball background subtraction. For HigA or HigBA bound to either higO1 or higO2, the hig DNA were fit using a site-specific binding equation in GraphPad (Fraction bound=

bound

free+bound).

AcknowledgementsThis work was supported in part by a National Science Foundation CAREER award MCB 0953714 (CMD), a National Institutes of Health (NIH) Biochemistry, Cellular and Developmental Biology Graduate Training Grant (5T32GM8367) and a NIH National Research Service Award Fellowship GM108351 (MAS). CMD is a Burroughs Wellcome

Fund Investigator in the Pathogenesis of Infectious Disease. We thank F. M. Murphy IV and staff members of the NE-CAT beamlines for assistance during data collection, Dr. G. Conn for critical reading of the manuscript and Dunham lab mem-ber S. Miles for technical assistance. This work is based upon research conducted at the NE-CAT beamlines, which are funded by the NIGMS from the NIH (P41 GM103403), and at the SER-CAT beamline. The Pilatus 6M detector on 24-ID-C beam line is funded by a NIH-ORIP HEI grant (S10 RR029205). This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357.

Author contributions

MAS and JM, conception and design, acquisition of data, analysis and interpretation of data, drafting and revis-ing the article. EH, analysis and interpretation of data, revising the article. PL, acquisition of data, analysis and interpretation of data and revising the article. CMD, con-ception and design, interpretation of data, drafting and revising the article. DW, NI and SC, acquisition of data.

Data depositionCrystallography, atomic coordinates and structure factors have been deposited in the Protein Data Bank, www.pdb.org (PDB codes 6CF1, 6CHV).

References

Adams, P.D., Afonine, P.V., Bunkoczi, G., Chen, V.B., Davis, I.W., Echols, N., et al. (2010) PHENIX: a comprehensive Python-based system for macromolecular structure solu-tion. Acta Crystallographica, 66, 213–221.

Afif, H., Allali, N., Couturier, M. and Van Melderen, L. (2001) The ratio between CcdA and CcdB modulates the tran-scriptional repression of the ccd poison-antidote system. Molecular Microbiology, 41, 73–82.

Blanchet, C., Pasi, M., Zakrzewska, K. and Lavery, R. (2011) CURVES+ web server for analyzing and visual-izing the helical, backbone and groove parameters of nucleic acid structures. Nucleic Acids Research, 39, W68–W73.

Boggild, A., Sofos, N., Andersen, K.R., Feddersen, A., Easter, A.D., Passmore, L.A. and Brodersen, D.E. (2012) The crystal structure of the intact E. coli RelBE toxin-anti-toxin complex provides the structural basis for conditional cooperativity. Structure, 20, 1641–1648.

Brown, B.L., Grigoriu, S., Kim, Y., Arruda, J.M., Davenport, A., Wood, T.K., et al. (2009) Three dimensional structure of the MqsR:MqsA complex: a novel TA pair comprised of a toxin homologous to RelE and an antitoxin with unique properties. PLoS Pathogens, 5, e1000706.

Brown, B.L., Lord, D.M., Grigoriu, S., Peti, W. and Page, R. (2013) The Escherichia coli toxin MqsR destabilizes the

http://www.pdb.org

http://www.pdb.org



transcriptional repression complex formed between the antitoxin MqsA and the mqsRA operon promoter. Journal of Biological Chemistry, 288, 1286–1294.

Christensen, S.K., Maenhaut-Michel, G., Mine, N., Gottesman, S., Gerdes, K. and Van Melderen, L. (2004) Overproduction of the Lon protease triggers inhibition of translation in Escherichia coli: involvement of the yefM-yoeB toxin-anti-toxin system. Molecular Microbiology, 51, 1705–1717.

Christensen-Dalsgaard, M., Jorgensen, M.G. and Gerdes, K. (2010) Three new RelE-homologous mRNA interfer-ases of Escherichia coli differentially induced by environ-mental stresses. Molecular Microbiology, 75, 333–348.

Cole, J.L., Lary, J.W., Moody, T.P. and Laue, T.M. (2008) Analytical ultracentrifugation: sedimentation velocity and sedimentation equilibrium. Methods in Cell Biology, 84, 143–179.

Dam, J. and Schuck, P. (2004) Calculating sedimentation co-efficient distributions by direct modeling of sedimentation velocity concentration profiles. Methods in Enzymology, 384, 185–212.

De Jonge, N., Garcia-Pino, A., Buts, L., Haesaerts, S., Charlier, D., Zangger, K., et al. (2009) Rejuvenation of CcdB-poisoned gyrase by an intrinsically disordered pro-tein domain. Molecular Cell, 35, 154–163.

Dienemann, C., Boggild, A., Winther, K.S., Gerdes, K. and Brodersen, D.E. (2011) Crystal structure of the VapBC tox-in-antitoxin complex from Shigella flexneri reveals a hete-ro-octameric DNA-binding assembly. Journal of Molecular Biology, 414, 713–722.

Emsley, P., Lohkamp, B., Scott, W.G. and Cowtan, K. (2010) Features and development of Coot. Acta Crystallographica, 66, 486–501.

Garcia-Pino, A., Christensen-Dalsgaard, M., Wyns, L., Yarmolinsky, M., Magnuson, R.D., Gerdes, K. and Loris, R. (2008) Doc of prophage P1 is inhibited by its antitoxin partner Phd through fold complementation. Journal of Biological Chemistry, 283, 30821–30827.

Garcia-Pino, A., Balasubramanian, S., Wyns, L., Gazit, E., De Greve, H., Magnuson, R.D., et al. (2010) Allostery and intrinsic disorder mediate transcription regulation by con-ditional cooperativity. Cell, 142, 101–111.

Gerdes, K., Christensen, S.K. and Lobner-Olesen, A. (2005) Prokaryotic toxin-antitoxin stress response loci. Nature Reviews Microbiology, 3, 371–382.

Gonzalez Barrios, A.F., Zuo, R., Hashimoto, Y., Yang, L., Bentley, W.E. and Wood, T.K. (2006) Autoinducer 2 con-trols biofilm formation in Escherichia coli through a novel motility quorum-sensing regulator (MqsR, B3022). Journal of Bacteriology, 188, 305–316.

Guzman, L.M., Belin, D., Carson, M.J. and Beckwith, J. (1995) Tight regulation, modulation, and high-level ex-pression by vectors containing the arabinose PBAD pro-moter. Journal of Bacteriology, 177, 4121–4130.

Hadzi, S., Garcia-Pino, A., Haesaerts, S., Jurenas, D., Gerdes, K., Lah, J. and Loris, R. (2017) Ribosome-dependent Vibrio cholerae mRNAse HigB2 is regulated by a beta-strand sliding mechanism. Nucleic Acids Research, 45, 4972–4983.

Harrison, J.J., Wade, W.D., Akierman, S., Vacchi-Suzzi, C., Stremick, C.A., Turner, R.J. and Ceri, H. (2009) The chromosomal toxin gene yafQ is a determinant of

multidrug tolerance for Escherichia coli growing in a biofilm. Antimicrobial Agents and Chemotherapy, 53, 2253–2258.

Holm, L. and Rosenstrom, P. (2010) Dali server: conser-vation mapping in 3D. Nucleic Acids Research, 38, W545–W549.

Hurley, J.M. and Woychik, N.A. (2009) Bacterial toxin HigB associates with ribosomes and mediates translation-de-pendent mRNA cleavage at A-rich sites. Journal of Biological Chemistry, 284, 18605–18613

Kabsch, W. (2010) Xds. Acta Crystallographica, 66, 125–132.Kamada, K. and Hanaoka, F. (2005) Conformational change

in the catalytic site of the ribonuclease YoeB toxin by YefM antitoxin. Molecular Cell, 19, 497–509.

Kamada, K., Hanaoka, F. and Burley, S.K. (2003) Crystal structure of the MazE/MazF complex: molecular bases of antidote-toxin recognition. Molecular Cell, 11, 875–884.

Kedzierska, B., Lian, L.Y. and Hayes, F. (2007) Toxin-antitoxin regulation: bimodal interaction of YefM-YoeB with paired DNA palindromes exerts transcriptional au-torepression. Nucleic Acids Research, 35, 325–339.

Kirkpatrick, C.L., Martins, D., Redder, P., Frandi, A., Mignolet, J., Chapalay, J.B., et al. (2016) Growth control switch by a DNA-damage-inducible toxin-antitoxin system in Caulobacter crescentus. Nature Microbiology, 1, 16008.

Lamoureux, J.S., Maynes, J.T. and Glover, J.N. (2004) Recognition of 5’-YpG-3’ sequences by coupled stacking/hydrogen bonding interactions with amino acid residues. Journal of Molecular Biology, 335, 399–408.

Li, G.Y., Zhang, Y., Inouye, M. and Ikura, M. (2008) Structural mechanism of transcriptional autorepression of the Escherichia coli RelB/RelE antitoxin/toxin module. Journal of Molecular Biology, 380, 107–119.

Li, G.W., Burkhardt, D., Gross, C. and Weissman, J.S. (2014) Quantifying absolute protein synthesis rates reveals prin-ciples underlying allocation of cellular resources. Cell, 157, 624–635.

Liu, Y., Zhang, X., Blumenthal, R.M. and Cheng, X. (2013) A common mode of recognition for methylated CpG. Trends in Biochemical Sciences, 38, 177–183.

Loris, R. and Garcia-Pino, A. (2014) Disorder- and dynam-ics-based regulatory mechanisms in toxin-antitoxin mod-ules. Chemical Reviews, 114, 6933–6947.

Magnuson, R. and Yarmolinsky, M.B. (1998) Corepression of the P1 addiction operon by Phd and Doc. Journal of Bacteriology, 180, 6342–6351.

Moyed, H.S. and Bertrand, K.P. (1983) hipA, a newly recog-nized gene of Escherichia coli K-12 that affects frequency of persistence after inhibition of murein synthesis. Journal of Bacteriology, 155, 768–775.

Overgaard, M., Borch, J., Jorgensen, M.G. and Gerdes, K. (2008) Messenger RNA interferase RelE controls relBE transcription by conditional cooperativity. Molecular Microbiology, 69, 841–857.

Page, R. and Peti, W. (2016) Toxin-antitoxin systems in bacterial growth arrest and persistence. Nature Chemical Biology, 12, 208–214.

Ruangprasert, A., Maehigashi, T., Miles, S.J., Giridharan, N., Liu, J.X. and Dunham, C.M. (2014) Mechanisms of toxin inhibition and transcriptional repression by Escherichia coli DinJ-YafQ. Journal of Biological Chemistry, 289, 20559–20569.



Schrodinger, L.L.C. (2010) The PyMOL molecular graph-ics system, Version, 2.0 Schrödinger, LLC. Available at: https://pymol.org/2/support.html?

Schuck, P. (2000) Size-distribution analysis of macromol-ecules by sedimentation velocity ultracentrifugation and lamm equation modeling. Biophysical Journal, 78, 1606–1619.

Schuck, P.Z.H., Brautigam, C.A. and Ghirlando, R. (2016) Basic Principles of Analytical Ultracentrifugation. Boca Raton: CRC Press.

Schumacher, M.A., Piro, K.M., Xu, W., Hansen, S., Lewis, K. and Brennan, R.G. (2009) Molecular mechanisms of HipA-mediated multidrug tolerance and its neutralization by HipB. Science, 323, 396–401.

Schureck, M.A., Maehigashi, T., Miles, S.J., Marquez, J., Cho, S.E., Erdman, R. and Dunham, C.M. (2014) Structure of the Proteus vulgaris HigB-(HigA)2-HigB toxin-antitoxin complex. Journal of Biological Chemistry, 289, 1060–1070.

Schureck, M.A., Dunkle, J.A., Maehigashi, T., Miles, S.J. and Dunham, C.M. (2015) Defining the mRNA recognition signature of a bacterial toxin protein. Proceedings of the National Academy of Sciences, 112, 13862–13867.

Schureck, M.A., Maehigashi, T., Miles, S.J., Marquez, J. and Dunham, C.M. (2016a) mRNA bound to the 30S subunit is a HigB toxin substrate. RNA, 22, 1261–1270.

Schureck, M.A., Repack, A., Miles, S.J., Marquez, J. and Dunham, C.M. (2016b) Mechanism of endonuclease cleavage by the HigB toxin. Nucleic Acids Research, 44, 7944–7953.

Shimon, L.J. and Harrison, S.C. (1993) The phage 434 OR2/R1-69 complex at 2.5 Å resolution. Journal of Molecular Biology, 232, 826–838.

Tian, Q.B., Ohnishi, M., Tabuchi, A. and Terawaki, Y. (1996) A new plasmid-encoded proteic killer gene system: cloning,

sequencing, and analyzing hig locus of plasmid Rts1. Biochemical and Biophysical Research Communications, 220, 280–284.

Tian, Q.B., Ohnishi, M., Murata, T., Nakayama, K., Terawaki, Y. and Hayashi, T. (2001) Specific protein-DNA and pro-tein-protein interaction in the hig gene system, a plas-mid-borne proteic killer gene system of plasmid Rts1. Plasmid, 45, 63–74.

Turnbull, K.J. and Gerdes, K. (2017) HicA toxin of Escherichia coli derepresses hicAB transcription to se-lectively produce HicB antitoxin. Molecular Microbiology, 104, 781–792.

Van Melderen, L., Bernard, P. and Couturier, M. (1994) Lon-dependent proteolysis of CcdA is the key control for activation of CcdB in plasmid-free segregant bacteria. Molecular Microbiology, 11, 1151–1157.

Wang, X., Kim, Y., Hong, S.H., Ma, Q., Brown, B.L., Pu, M., et al. (2011) Antitoxin MqsA helps mediate the bacterial gen-eral stress response. Nature Chemical Biology, 7, 359–366.

Watkins, D., Hsiao, C., Woods, K.K., Koudelka, G.B. and Williams, L.D. (2008) P22 c2 repressor-operator complex: mechanisms of direct and indirect readout. Biochemistry, 47, 2325–2338.

Zhao, H., Lomash, S., Glasser, C., Mayer, M.L. and Schuck, P. (2013) Analysis of high affinity self-association by flu-orescence optical sedimentation velocity analytical ul-tracentrifugation of labeled proteins: opportunities and limitations. PLoS ONE, 8, e83439.

Supporting InformationAdditional supporting information may be found online in the Supporting Information section at the end of the article

https://pymol.org/2/support.html?

Documents

Structural basis of transcriptional regulation by the HigA